SYSTEM AND METHOD FOR GENERATING RECOMMENDATIONS FOR THE DISTRIBUTION OF VEHICLE CARGO

Abstract

System in a vehicle (100), wherein the system comprises: a memory that stores computer-readable instructions; and one or more processors configured to execute the computer-readable instructions, wherein the computer-readable instructions control the one or more processors to execute processes that include: virtual division of a loading area (120) of the vehicle (100) into several virtual cells (130); Determine whether the loading area (120) includes one or more weight sensors (150); Generating by implementing machine learning of a load profile for the loading platform (120) by obtaining weight values ​​from each of the one or more weight sensors (150) based on determining that the loading platform (120) includes the one or more weight sensors (150), and obtaining weight values ​​from sensor measurements from at least one of one or more tire pressure sensors and one or more suspension sensors (190), wherein the load profile provides a weight estimate in each of the multiple virtual cells (130) based on a load on the loading platform (120); Determining a first total weight estimate as the sum of the weight estimates of the virtual cells (1-16, 130) of the left loading area half and a second total weight estimate as the sum of the weight estimates of the virtual cells (17-32, 130) of the right loading area half; Determining a sigmoid function of a difference between the first and second total weight estimates, where the sigmoid function is defined as sig ( Δ ll − r ) − 1 1 + e − Δ ll − r and represents a sigmoid curve, where Δl l-r expresses the difference in the total weight estimates; and If the left side of the sigmoid curve is heavier than the right side of the sigmoid curve by more than a predetermined or learned threshold, provide a recommendation to move part of the load within the loading area (120) to the right in order to balance the load on the loading area (120).

Description

INTRODUCTION

[0001] The present invention relates to a system and a method for generating recommendations for action in order to arrange vehicle loads as evenly distributed as possible.

[0002] In this regard, for background information, reference should first be made to US 2021 / 0158185A1, from which such a procedure is described in this way.

[0003] Vehicles (e.g., cars, trucks, construction equipment, agricultural machinery) can be used to transport heavy loads. Generally, a manufacturer specifies the weight that can be safely transported by a given type of vehicle. The permissible axle load (GAWR) for a given vehicle, for example, indicates the maximum distributed weight that can be borne by one axle of the vehicle. Transporting a load exceeding the recommended weight can have several adverse effects, including vehicle instability. Therefore, it is desirable to establish guidelines for determining and distributing vehicle load weight. SUMMARY

[0004] According to the invention, a system in a vehicle is presented which is characterized by the features of claim 1.

[0005] In addition to one or more of the features described here, obtaining the weight values ​​from the sensor measurements includes implementing machine learning to map the sensor measurements to the weight values.

[0006] In addition to one or more of the features described here, the processes also include merging the weight values ​​obtained from two or more types of sensor measurements.

[0007] In addition to one or more of the features described here, the processes also include obtaining images of the loading area from a camera and determining which of the virtual cells are occupied, with the generation of the load profile involving the use of the occupancy of the virtual cells during the implementation of machine learning.

[0008] In addition to one or more of the features described here, providing the instructions includes providing language.

[0009] In addition to one or more of the features described here, providing the instructions includes providing text.

[0010] Furthermore, according to the invention, a method is presented which is characterized by the features of claim 4.

[0011] In addition to one or more of the features described here, obtaining the weight values ​​from the sensor measurements includes implementing machine learning to map the sensor measurements to the weight values.

[0012] In addition to one or more of the features described here, the method also includes merging the weight values ​​obtained from two or more types of sensor measurements.

[0013] In addition to one or more of the features described here, generating the load profile includes assigning the weight values ​​of the weight estimate in each of the multiple virtual cells.

[0014] In addition to one or more of the features described here, the method also includes obtaining images of the loading area from a camera and determining which of the virtual cells are occupied, with the generation of the load profile including using the occupancy of the virtual cells during the implementation of machine learning.

[0015] In addition to one or more of the features described here, providing the instructions includes providing language.

[0016] In addition to one or more of the features described here, providing the instructions includes providing text.

[0017] The above features and advantages and other features and advantages of the invention are readily apparent from the following detailed description when considered in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Further features, advantages and details appear in the following detailed description only as examples, the detailed description referring to the drawings; they show: Fig. 1 a block diagram of a vehicle which includes a recommendation for action for determining and distributing the load weight according to one or more embodiments; Fig. 2 a process flow of a method for providing a recommendation for action for determining and distributing the load weight in a vehicle according to one or more embodiments; Fig. 3. Detailed aspects of generating a load profile as part of the recommendation for determining and distributing the load weight according to one or more embodiments; and Fig. 4 detailed processes involved in providing the recommendation for action according to one or more embodiments. DETAILED DESCRIPTION

[0019] The following description is merely exemplary. It should be recognized that throughout the drawings, corresponding reference symbols indicate the same or corresponding parts and features.

[0020] The embodiments of the systems and methods described in detail here relate to a recommendation for determining and distributing the vehicle's load weight. As previously noted, transporting a load that exceeds the maximum recommended weight for a given vehicle can have adverse effects. According to one or more embodiments, a warning can be issued to the driver when it is determined that the load weight is approaching or exceeding a maximum. Even if a load does not exceed the recommended maximum weight, its distribution within the vehicle may be such that stability is compromised, or one or more tires may bear a larger proportion of the load and, for example, experience increased wear. Furthermore, an unbalanced load can interfere with the calibration of sensors, such as...a camera, by causing the vehicle to tilt from its reference position. According to one or more embodiments, these problems can be avoided based on the distribution action recommendation, as described in detail.

[0021] According to an exemplary embodiment, Fig. 1. A block diagram of a vehicle 100, which includes a recommendation for determining and distributing the load weight. The in Fig. The exemplary vehicle 100 shown is a flatbed truck 101 with a flatbed as a loading platform 120. This exemplary illustration is not intended to restrict the type of vehicle 100 or the type of loading platform 120. A controller 110 of the vehicle 100 can determine the load weight on the loading platform 120 and provide a recommendation for load distribution, as described in detail below.

[0022] The controller 110 may include a processing circuit arrangement comprising an application-specific integrated circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or grouped), and memory that executes one or more software or firmware programs, a combinational logic circuit, and / or other suitable components that provide the described functionality. The controller 110 may also include communication systems that facilitate communication with devices 105 (e.g., a smartphone, a tablet) that may be worn, for example, by an occupant of the vehicle 100.The memory of the controller 110 can contain a non-transient computer-readable medium that stores instructions which, when processed by one or more processors of the controller 110, implement a method for executing the action recommendation for determining and distributing the load weight in a vehicle 100 according to one or more embodiments described in detail herein.

[0023] The center of gravity (CoG) of the loading area 120 is indicated, with the loading area 120 shown divided into virtual cells 130 forming a virtual grid. The CoG is the point where the weight is evenly distributed and the point where the entire weight can be considered concentrated on the loading area 120. The virtual cells 130 are numbered for explanatory purposes. The controller 110 estimates the weight of the load in each virtual cell 130 to provide the recommendation for load distribution. Based on the size of the loading area 120 and the resolution at which load distribution information is desired or required, fewer or more virtual cells 130 may be used. The recommendation for action provided by the controller 110 can be visual (e.g., text displayed on a device 105 or an infotainment display, a color display, or other display), audible (e.g., a warning signal), orVoice commands via a loudspeaker 195 based on text-to-speech processing or predefined commands), haptic or via any other available feedback mechanism (commonly referred to as the human-machine interface (HMI)) of the vehicle 100.

[0024] The vehicle 100 can contain any number of sensors 140 that obtain information about the vehicle 100 and about objects near the vehicle 100. One type of sensor 140 that is in Fig. Figure 1 shows a weight sensor 150. Four optional weight sensors 150 are shown at the four corners of the loading platform 120. Another type of sensor 140, which is in Fig. As shown in Figure 1, there is a camera 160. While additional forward-, side-, and rear-facing cameras 160 can be used to detect objects near the vehicle 100, the one shown in Figure 1 is used for the camera 160. Fig. 1 camera shown, 160 images of the loading area, 120. Additional types of vehicle sensors, 140, which are relevant for discussion regarding Fig. 2. The relevant components are the tire pressure monitoring system (TPMS) 170, the inertial measurement unit (IMU) 180 and the suspension sensors 190. Fig. Section 2 describes in detail the processes by which the controller 110 estimates the weight in each virtual cell 130 using these sensors 140.

[0025] Fig. 2 is a process flow of a method 200 for providing the recommendation for action for determining and distributing the load weight in a vehicle 100 according to one or more embodiments. The in Fig. The processes shown in Figure 2 can be executed by the controller 110 based on information from various sensors 140. Block 210 performs a check to determine whether a load has been placed on the loading platform 120. This check can be performed, for example, based on the camera 160. The determination, based on the check in block 210, that a load is present on the loading platform 120 triggers the remaining processes. If the check in block 210 indicates that there is no load on the loading platform 120, then the check is repeated as shown.

[0026] This check in block 210 can be repeated periodically at a predefined time interval or based on a predefined event. For example, if a key fob corresponding to vehicle 100 is detected, the check in block 210 can be performed periodically until vehicle 100 moves. The check can be repeated, for example, every time vehicle 100 is parked. The trigger for the check in block 210 is not intended to be restricted by the examples. If the check in block 210 indicates that there is a load on loading platform 120, then the processes in blocks 220 and 260 are triggered. In block 260, the generation of a load profile is based on information received and provided by other processes, as described in detail. In block 220, a check is performed to determine whether there are weight sensors 150 on loading platform 120.

[0027] If the check in block 220 indicates that there are 120 weight sensors 150 on the loading platform, then block 230 obtains a weight measurement from each of the weight sensors 150 on the loading platform. Block 235 then performs a check to see if the weight sensors 150 indicate that a maximum weight has been exceeded. The maximum weight of a load that can be transported by the vehicle 100 can be set, for example, by the manufacturer to ensure stability. If the maximum weight has been exceeded according to the check in block 235, then a warning is issued in block 237. The warning can be issued via any available HMI or mechanism (e.g., loudspeaker 195).

[0028] If the maximum weight has not been exceeded according to the test in block 235, then the load profile is generated in block 260. According to exemplary embodiments, the processing to obtain a load profile in block 260 can continue after a warning has been issued in block 237, even if the maximum weight has been exceeded according to the test in block 235. Additional tests can be performed even after processing in block 260 has been completed. For example, if the weight is within a certain percentage (e.g., 90 percent) of the maximum weight, a warning can still be issued. This warning can also be provided via visual, audible, or haptic output.

[0029] If the check in block 220 indicates that there are no weight sensors 150 on the loading platform 120, then the weight information from other sensors must be gathered in block 240. The processes in block 240 involve the TPMS 170, the IMU 180, and the suspension sensors 190. As regarding Fig. As further discussed in section 3, the information from each of these sensors (140) is used to assign, for example, weight, using machine learning.

[0030] In block 250, obtaining an occupancy mapping of the virtual grid refers to using the images from camera 160 to estimate, for example, the parts (e.g., virtual cells 130) of the loading area 120 that are occupied by the load detected in block 210. The images from camera 160 can be superimposed on the virtual cells 130, whereby, for example, a higher weight can be given to virtual cells 130 that, according to the images, contain a load. This weighting can be used in block 310 to obtain estimates of the weight in each virtual cell 130, as described above. Fig. 3 will be discussed further.

[0031] Based on the trigger from block 210 and the information from blocks 250 and 230 or 240 (based on whether there are weight sensors 150 according to the check in block 220), the processes include generating a load profile in block 260. This process is carried out with regard to Fig. 3 further discussed and leads to an estimated weight in each virtual cell 130. In block 270, providing a recommendation for loading refers to the controller 110, which provides suggestions on how to move the load to better balance the weight on the loading platform 120, as with regard to Fig. 4 is described in detail. As previously noted, the action recommendation can be provided via a device 105 with which the controller 110 communicates, via an indoor or outdoor loudspeaker 195, or via any other known HMI or mechanism.

[0032] Fig. Section 3 describes in detail the aspects of generating a load profile in block 260. As with regard to Fig. As discussed in section 2, the weights in block 230 can be obtained from the weight sensors 150, or in block 240 based on other sensors 140. In block 230, a weight value can be obtained from each of the weight sensors 150 on the loading platform 120 (e.g., four weight values ​​according to the exemplary illustration in [reference]). Fig. 1) In block 240, a weight value can be obtained from each tire pressure sensor of the TPMS 170, from each suspension sensor 190, and from the IMU 180. The weight values ​​can be obtained based on a mapping of the sensor measurement to the weight. The measurement taken by each of the four sensors of the in Fig. The tire pressure measured by the TPMS 170 shown can be assigned to, for example, four weight values. The assignment can be different for each type of sensor 140 (e.g., a different assignment for the TPMS 170 and the IMU 180) and can be a machine learning-based assignment implemented by the controller 110 after training.

[0033] Generating the load profile in block 260 includes the process in block 310 and, if sensors 140 other than the weight sensors 150 are used, also the process in block 320. In block 310, estimating the weight in each virtual cell 130 involves mapping the weight values. The weight values ​​can be obtained either in block 230 from each of the weight sensors 150 or in block 240 from the other sensors 140 (e.g., the TPMS 170, the IMU 180, the suspension sensors 190). Like the mapping in block 240, the mapping of the weight value to a weight estimate in each virtual cell 130 can be implemented using machine learning, and it can differ for each type of sensor 140. This mapping can additionally use the occupancy mapping of the virtual grid from block 250.

[0034] The one in block 230 of each of the four in Fig. The weight values ​​obtained from the 150 weight sensors shown can, for example, be used for weight estimates for each of the thirty-two in Fig. The virtual cells 130 shown in block 1 can be assigned. For example, if the two weight sensors 150 on the left side of the vehicle 100 measure higher weight values ​​than the two weight sensors 150 on the right side, the occupancy assignment of the virtual grid from block 250 can be used to determine whether the load on the loading platform 120 is distributed more between the front or the rear virtual cells 130 on the left side of the vehicle 100. That is, the occupancy assignment of the virtual grid from block 250 can be added as a weight to the assignment process.

[0035] The mapping process in block 310 provides the load profile (i.e., the weight estimate for each virtual cell 130) if the weight values ​​in block 230 are obtained from the weight sensors 150. If the weight values ​​in block 240 are obtained from the other sensors 140, a set of weight estimates is obtained for each type of sensor 140 for each of the virtual cells 130. That is, in the example case, three separate load profiles are obtained based on the three sensors 140 (i.e., the TPMS 170, the IMU 180, and the suspension sensors 190). Consequently, the merging process in block 320 is additionally required to generate the load profile.

[0036] In block 320, the load profiles obtained from the various sensors 140 used in block 240 are merged. Each weight estimate for each virtual cell 130 is obtained by assigning a confidence score, which can be used as a weight. For each virtual cell 130 (e.g., i is an index of the virtual cells 130 and i = 1 to 32 in the example according to Fig. 1) is therefore the weight estimate obtained in block 310 using the TPMS 170 l Ti with a variance σ Ti , is the weight estimate obtained using the IMU 180 l li with a variance σ Ii and is the weight estimate obtained using the suspension sensors 190 l Si with a variance σ Si A corresponding weighting (w Ti , w Ii and w Si ), which each weight estimate (l Ti , l Ii and l Si) is inversely proportional to the variance (σ) Ti , σ Ii and σ Si ) of the sampling modality (i.e., of the associated sensor 140). That is, the more secure modality is given a higher weight during an information fusion process. The fused estimated weight in each virtual cell 130 is given by: li=median(wTiwTi+wli+wSilTi+wIiwTi+wIi+wSilIi+wSiwTi+wIi+wSilSi)

[0037] Fig. Section 4 explains in detail the processes involved in providing the recommendation for action in block 270. Fig. 2 are included. The processes are based on the load profile generated in block 260. In block 410, a check is performed to determine whether the load on the left side of the loading platform 120 is heavier than the load on the right side by a predefined threshold value ε. The value of the predefined threshold value ε could, for example, be an average human weight used in vehicles. In the Fig. In the exemplary case shown, the test in block 410 may include summing the weight estimates (according to the load profile) for the virtual cells 130 numbered 1-16 and subtracting a sum of the weight estimates for the virtual cells 130 numbered 17-32.

[0038] This difference in the weight estimates of the left and right sides can be expressed as Δl l-r can be expressed. Alternatively, a sigmoid function of the difference in block 410 can be expressed as: sig(Δll−r)=11e−Δll−r The result of Eq. 2 is a sigmoid curve with a range (0, 1). If the left side is heavier than the right side by more than the predetermined or learned threshold ε (e.g., a predetermined value that is an average human weight) (e.g., Δl) l-r > ε or sig(Δl l-r If the value is greater than 0.5, then block 420 provides a recommendation to move some of the load towards the right side. As previously noted, the recommendation can be provided via loudspeaker 195, a device 105, or more than one output.

[0039] If the test in block 410 indicates that the load on the left side of loading platform 120 is not heavier than the load on the right side by the specified threshold value ε, then the opposite scenario is tested in block 430. Specifically, block 430 performs a test to determine whether the load on the right side of loading platform 120 is heavier than the load on the left side by the specified threshold value ε. In the Fig. In the exemplary case shown, the test in block 430 may include summing the weight estimates (according to the load profile) for the virtual cells 130 numbered 17-32 and subtracting a sum of the weight estimates for the virtual cells 130 numbered 1-16.

[0040] This difference in the weight estimates of the left and right sides can be expressed as Δl r-lcan be expressed. Alternatively, in block 430, a sigmoid function of the difference can be expressed by modifying equation 2 by adding Δl. r-l instead of Δl l-r to use, will be obtained. If the right side is heavier than the left side by more than the specified threshold ε (e.g., Δl) r-l > ε or sig(Δl r-l If the value of the load on the right side of loading platform 120 is greater than the specified threshold ε, then block 440 provides a recommendation to move some of the load towards the left side. If, instead, the test in block 430 indicates that the load on the right side of loading platform 120 is not heavier than the load on the left side by the specified threshold ε, then no recommendation is issued according to block 450.

[0041] Block 460 performs a test to determine whether the load in the front part of the loading area 120 is heavier than the load in the rear part by a predetermined threshold value ε. In the Fig. In the exemplary case shown, the test in block 460 may include summing the weight estimates (according to the load profile) for the virtual cells 130 numbered 5-8, 13-16, 21-24 and 29-32 and subtracting a sum of the weight estimates for the virtual cells 130 numbered 1-4, 9-12, 17-20 and 16-28.

[0042] This difference between the weight estimates of the front part and the rear part can be expressed as Δl f-b can be expressed. Alternatively, in block 460, a sigmoid function of the difference can be expressed by modifying equation 2 by Δl. f-b instead of Δl l-r to use, will be obtained. If the front part is heavier than the rear part by more than the specified threshold ε (e.g. Δl) f-b > ε or sig(Δl f-bIf the value of the load on the right side of the loading platform is greater than 0.5, then block 470 provides a recommendation to move some of the load towards the rear. If, instead, the test in block 430 indicates that the load on the right side of the loading platform 120 is not heavier than the load on the left side by the specified threshold value ε, then no recommendation is issued according to block 450.

[0043] If the test in block 460 indicates that the load in the front part of the loading area 120 is not heavier than the load in the rear part by the specified threshold value ε, then the opposite scenario is tested in block 480. Specifically, block 480 performs a test to determine whether the load in the rear part of the loading area 120 is heavier than the load in the front part by the specified threshold value ε. In the Fig.In the exemplary case shown, the test in block 410 may include summing the weight estimates (according to the load profile) for the virtual cells 130 numbered 1-4, 9-12, 17-20 and 16-28 and subtracting a sum of the weight estimates for the virtual cells 130 numbered 5-8, 13-16, 21-24 and 29-32.

[0044] This difference between the weight estimates of the rear part and the front part can be expressed as Δl b-f can be expressed. Alternatively, in block 480, a sigmoid function of the difference can be expressed by modifying equation 2 to account for Δl. b-f instead of Δl l-r to be used, obtained. If the rear part is heavier than the front part by more than the specified threshold ε (e.g. Δl) b-f > ε or sig(Δl b-fIf the value of the load in the rear of the loading area is greater than 0.5, then block 490 provides a recommendation to move some of the load towards the front. If, instead, the test in block 480 indicates that the load in the rear of the loading area is not heavier than the load in the front by the specified threshold value ε, then no recommendation is issued according to block 450.

[0045] Once the load has been moved based on the action recommendation provided by the processes in Block 270, the processes in Blocks 220 to 270 can then be repeated to determine whether an additional action recommendation and adjustment of the load are required. The processes can be triggered and repeated as often as necessary when loads are added, removed, or moved. According to one or more embodiments, the load profile-based action recommendation for weight distribution promotes load balancing that is either impractical or impossible to perform manually. This is because load balancing to optimize stability and other factors becomes much more complicated when multiple items of different weights are arranged on the loading platform 120, whereas a single item can be positioned on the loading platform 120 (e.g., at the center of gravity).

Claims

[1] System in a vehicle (100), wherein the system comprises: a memory that stores computer-readable instructions; and one or more processors configured to execute the computer-readable instructions, wherein the computer-readable instructions control the one or more processors to execute processes that include: virtual division of a loading area (120) of the vehicle (100) into several virtual cells (130); Determine whether the loading area (120) includes one or more weight sensors (150); Generating by implementing machine learning of a load profile for the loading platform (120) by obtaining weight values ​​from each of the one or more weight sensors (150) based on determining that the loading platform (120) includes the one or more weight sensors (150), and obtaining weight values ​​from sensor measurements from at least one of one or more tire pressure sensors and one or more suspension sensors (190), wherein the load profile provides a weight estimate in each of the multiple virtual cells (130) based on a load on the loading platform (120); Determining a first total weight estimate as the sum of the weight estimates of the virtual cells (1-16, 130) of the left loading area half and a second total weight estimate as the sum of the weight estimates of the virtual cells (17-32, 130) of the right loading area half; Determining a sigmoid function of a difference between the first and second total weight estimates, where the sigmoid function is defined as sig(Δll−r)−11+e−Δll−r and represents a sigmoid curve, where Δl l-r expresses the difference in the total weight estimates; and If the left side of the sigmoid curve is heavier than the right side of the sigmoid curve by more than a predetermined or learned threshold, provide a recommendation to move part of the load within the loading area (120) to the right in order to balance the load on the loading area (120). [2] System according to claim 1, wherein the processes further comprise obtaining images of the loading area (120) from a camera (160) and determining which of the virtual cells (130) are occupied, and generating the load profile, using the occupancy of the virtual cells (130) during the implementation of machine learning. [3] System according to claim 1, wherein providing the action recommendation comprises providing speech or providing text. [4] Procedure that includes: virtual division using a processor of a loading area (120) of a vehicle (100) into several virtual cells (130); Determine whether the loading area (120) includes one or more weight sensors (150); Generating by implementing machine learning of a load profile for the loading platform (120) by obtaining weight values ​​from each of the one or more weight sensors (150) based on determining that the loading platform (120) includes the one or more weight sensors (150), and obtaining weight values ​​from sensor measurements from at least one of one or more tire pressure sensors and one or more suspension sensors (190), wherein the load profile provides a weight estimate in each of the multiple virtual cells (130) based on a load on the loading platform (120); Determining a first total weight estimate as the sum of the weight estimates of the virtual cells (1-16, 130) of the left loading area half and a second total weight estimate as the sum of the weight estimates of the virtual cells (17-32, 130) of the right loading area half; Determining a sigmoid function of a difference between the first and second total weight estimates, where the sigmoid function is defined as sig(Δll−r)−11+e−Δll−r and represents a sigmoid curve, where Δl l-r expresses the difference in the total weight estimates; and If the left side of the sigmoid curve is heavier than the right side of the sigmoid curve by more than a predetermined or learned threshold, provide a recommendation to move part of the load within the loading area (120) to the right in order to balance the load on the loading area (120). [5] Method according to claim 4, which further comprises obtaining images of the loading area (120) from a camera (160) and determining which of the virtual cells (130) are occupied, and includes generating the load profile using the occupancy of the virtual cells (130) during the implementation of machine learning. [6] Method according to claim 4, wherein providing the recommendation for action includes providing language or providing text.

Images